feat(etl): 实现 hk_songs_test 读取与 hikoon-data 关联

+from typing import Iterator
+import pymysql
+from .config import PLATFORMS
+
+
+_HK_SONGS_QUERY = """
+SELECT id, name, lyricist, composer, audio_url, lyrics_url,
+       cover_url, singer, issue_time, source_song_id
+FROM hk_songs_test
+WHERE deleted = '0'
+  AND name IS NOT NULL AND name != ''
+  AND audio_url IS NOT NULL AND audio_url != ''
+  AND singer IS NOT NULL AND singer != ''
+ORDER BY id
+LIMIT %s OFFSET %s
+"""
+
+_PLATFORM_QUERY = """
+SELECT
+    sar.song_id AS source_song_id,
+    mr.platform,
+    mr.platform_unique_key,
+    mr.platform_mid,
+    mr.album_audio_id,
+    sar.is_main_version
+FROM hk_song_and_record sar
+JOIN hk_music_record mr ON mr.id = sar.record_id
+WHERE sar.song_id IN ({placeholders})
+  AND mr.platform IN ('1','2','4')
+  AND mr.platform_unique_key IS NOT NULL
+  AND mr.platform_unique_key != ''
+  AND mr.deleted = 0
+ORDER BY sar.song_id, mr.platform, sar.is_main_version DESC
+"""
+
+
+def iter_hk_songs_batches(conn: pymysql.Connection, batch_size: int) -> Iterator[list[dict]]:
+    offset = 0
+    with conn.cursor() as cur:
+        while True:
+            cur.execute(_HK_SONGS_QUERY, (batch_size, offset))
+            rows = cur.fetchall()
+            if not rows:
+                break
+            yield rows
+            if len(rows) < batch_size:
+                break
+            offset += batch_size
+
+
+def fetch_platform_records(source_conn: pymysql.Connection, song_ids: list[int]) -> list[dict]:
+    if not song_ids:
+        return []
+    placeholders = ','.join(['%s'] * len(song_ids))
+    query = _PLATFORM_QUERY.format(placeholders=placeholders)
+    with source_conn.cursor() as cur:
+        cur.execute(query, song_ids)
+        rows = cur.fetchall()
+
+    # 每个 (source_song_id, platform) 只保留 is_main_version 最高的一条
+    seen = {}
+    for row in rows:
+        key = (row['source_song_id'], row['platform'])
+        if key not in seen:
+            seen[key] = row
+    return list(seen.values())